JPH07117961B2 - Document data registration method - Google Patents

Description

Detailed Description of the Invention [Industrial applications]

The present invention prints on paper in a literature database or
The present invention relates to a document data registration method for automatically registering bibliographic items from written documents.

[Prior art]

For the conventional method of extracting bibliographic items from documents printed on paper, see the 8th International Conference on Pattern Recognition (1986).
745 to 748 (Proc.8th Int.Conf.Pattern Recogn
ition, pp.745-748 (1986)). Here, it is described that a character string such as an author name and an affiliated institution is extracted from image data from a document.

[Problems to be Solved by the Invention]

When the author name and the affiliated organization are taken out and registered in the reference database as attribute information for retrieval in the above-mentioned conventional technique, the notation of the author name and the affiliated organization is different and different depending on the target document. For this reason, if the author name and the institution belonging to the related art are registered as they are as the attribute information, there is a problem in that the relevant document cannot be searched from the document database, resulting in omission of the search. The present invention, even if the notation method of the author name or the institution described in the document printed on the paper varies depending on the document, it is possible to register the attribute information that enables the corresponding document to be searched without omission. It is also an object of the present invention to further provide a document database that analyzes a reference column of a document printed on paper and automatically associates the document with other documents.

[Means for Solving the Problems]

In order to achieve the above object, the extracted bibliographic items are transformed to generate variants, and the variants are used to search for desired documents. In addition, in order to associate a document with another document, a reference document described in the document is analyzed and bibliographic items are extracted to register the relationship between the documents. Further, the relevant documents are searched by setting the search condition from the generated variant.

[Action]

The bibliographic database generates variant words from the extracted bibliographic items. As a result, since the document related to the document of interest is registered together with the relevance, another document related to a certain document can be searched.

【Example】

An embodiment of the present invention will be described below with reference to FIG. The image input unit 100 is composed of an optical input device such as an image scanner, and the surface image of the document printed on the paper is
Image data is input page by page. The input image data is sent to the bibliographic item extraction unit 101, where the title, author name, institution, etc. and reference documents are separated and extracted as bibliographic items to be used for the search. Further, the variant word generation unit 102 generates variant words with different notations for the extracted bibliographic items. In the document file 103, not only the above-mentioned image data and the extracted bibliographic items are stored, but also the heteromorphic words generated for the bibliographic items are stored for use in the search. The search terminal 104 gives an instruction to search a large number of documents stored in the document file 103 using bibliographic items and variants thereof, and displays the search results on the screen. According to the present embodiment, bibliographic items are extracted from documents printed on paper, and a variant word having a different notation from the bibliographic items is generated, and a desired document is searched from a large number of documents using the bibliographic items. When doing so, there is an effect that the omission of the search is eliminated. FIG. 2 is a diagram for explaining a procedure for creating a search command for searching documents. First, in step 200, a user of the document database inputs a word for searching. Next, in step 201, an atypical word is generated for the input search word. Here, although the atypical word is generated for the search word input by the user, there is also a method of generating the atypical word in the literature database in advance. Further, in step 202, the generated variant word is displayed on the screen. This allows the user to correct or add to the generated variant. In step 203, the variant is used to generate a command file for searching the document database. Literature search will be executed using this command file. FIG. 3 is a diagram for explaining the generation of variants of the institution name. The institution name 300 is, for example, "Smith Kline and Fre
If it is "nch", a variant word generation unit 301 of the institution name
Then, an atypical word is generated for the institution name 300. The generated variants are shown at 302. Here, the original institution name is 300
Against the variant 303 “Smith, Kline French”, 304 “Smi
th, Kline & French '', 305 `` Smith, Kline, and Frenc
h '', 306 `` Smith, Kline, and French '', 307 `` Smith Klin
e, French ", 308" SK and F ", 309" SK & F ", 310" SK
F '', 311 `` Smith Kline '', 312 `` Smith Kline and Frenc
h ”and 313“ Smith Kline French ”are generated respectively. According to the present embodiment, even if the names of the institution belonging to the same institution are different, the variants of the institution name are generated. This has the effect of preventing omission of search. FIG. 4 is a diagram for explaining the generation of variants of the author name. When the author name 400 is, for example, “John Miles Smith”, the author name variant word generation unit 401 generates variant words for the author name 400. The generated variant is shown at 402. Here, for the original author name 400, the variant 403 “Miles Sm
ith, John ", 404" Smith, John ", 405" Smith, J. ", 406
"Smith, JM", 407 "M. Smith, J.", 408 "JMSmith",
409 "J. Smith", 410 "John Smith", 411 "John M. Smith
h ”are generated respectively. According to the present embodiment, even if the notation of the author name is different for the same author, a variant of the author's name is generated, so the omission of search due to the difference in notation method. There is an effect that can be prevented. FIG. 5 is an explanatory diagram of a procedure for registering the mutual relevance of documents. In step 500, a surface image is input for the document printed on the paper. Next, in step 501, the character is extracted and the character is recognized. In step 502, the arrangement of a character string having a character sequence is analyzed, and in step 503, bibliographic items are extracted from the character string. Further, a variant word is generated from the bibliographic items extracted in step 504, and in step 505, a relevant document search condition is generated using the variant word. In step 506, a document satisfying the search condition generated from a large number of documents is searched. And
In step 507, the document obtained in the previous step is registered as a related document. According to the present embodiment, it is possible to read the bibliographic items of the document printed on the paper and automatically register the relationship between the document and the document, so that the registration work in the document database can be automated and labor-saving. The effect is that you can do it. FIG. 6 is an explanatory diagram of a search procedure of the document database. In block 600, the bibliographic items are input, and in block 601, a variant word is created from the bibliographic items and a search command is generated. A search condition file is created from the search command generated in block 602, and a search is executed in block 604. This time,
A search execution instruction is given in block 603, and the relevant document is searched from the file 605 in which many documents are stored. The search result is stored in the search result file 606, and block 607 displays the search result. At this time, in block 608, the display format of the search result is instructed, and the result is displayed on the display 609. FIG. 7 is a diagram showing a procedure for extracting bibliographic items from a reference document. In step 700, enter the reference section. Here, the reference column is already extracted from the document printed on the paper for character recognition, and the reference column is a character code. In step 701, first, the blank is deleted from the reference literature section. Next, in step 702, the document number is retrieved. Then, in step 703, the author name is extracted.
In step 704, the title is extracted. Then step 7
The magazine name is extracted at 05, the volume number is extracted at step 706, and the page number is extracted at step 707. In addition, step 7
The issue year is extracted in 08, and the issue month is extracted in step 709. At step 710, the bibliographic items extracted from these reference columns are registered. According to the present embodiment, since the bibliographic items can be automatically extracted from the reference literature section, there is an effect that the work of giving the relation to other documents referred to in the reference literature section is automatically performed. FIG. 8 is a diagram showing a procedure of the blank deletion processing shown in step 701. The processes in and after step 801 are performed until the characters are exhausted in step 800. In step 801, a character is input for each digit, and in step 802, it is determined whether the input character is a blank character. If the character is a blank character, step 803 is performed. Here, if it is not a blank character, the process directly returns to step 801, and the character of the next digit is input. In step 803, it is determined whether the characters before and after the blank character satisfy any of the following conditions. The condition is that the space character is followed by "&", the space character is followed by "and", the space character is preceded by ".", Or the space character It is "," before, or """before the space character, or"("after the space character
That is, If this condition is not satisfied, the blank character is deleted in step 804 and the process returns to step 801. On the other hand, if this condition is satisfied, step 80
Return to 1. FIG. 9 is a diagram showing a procedure of extracting the document number shown in step 702. In step 900, the first digit character is input. Then, in step 901, the first digit is
It is determined whether or not the condition of “[”, “(”, or a number] is satisfied. If the above condition is satisfied, the process proceeds to step 902. If the condition is not satisfied, the extraction of the document number ends as unsuccessful in step 906. In step 902, the processes from step 903 onward are sequentially performed up to the n-th digit.
Is 15, for example. In step 903, the character of the corresponding digit is first input. Next, in step 904, it is determined whether or not the condition that the character of the corresponding digit is “]”, “)”, or “.” Is satisfied. If the above conditions are met, step 905
Then, the relevant digit is registered as the position of the end digit of the document number, and the process ends in step 907. On the other hand, if the condition is not satisfied in step 904, the process proceeds to step 903. Step 902
Repeat the process up to the nth digit with and if the condition is not satisfied,
In step 906, it is registered that the extraction of the document number is unsuccessful. FIG. 10 is a diagram for explaining the author name extraction procedure shown in step 703. In step 1000, it is determined whether or not the i digit satisfies the condition as the starting digit of the author name. The condition is that the i digit is the character digit that is the most recent character digit to the right of the last digit of the reference number. If this condition is satisfied, step 1001
The i digit is registered as the starting digit of the author name. In step 1002, it is determined whether or not the j digit satisfies the condition as the end digit of the author name. The condition is that (a) the j-th digit is a character and the j + 1-th digit is "", or
(B) j digit is a character and j + 1 and subsequent digits are ".:", Or (c) j digit is a character and j + 1 and subsequent digits are ". (" And is on the leftmost side. Or
(D) The j digit is a character, the j + 1 digit is “.”, The j + 2 digit is “,”, and the J + 3 digit is an uppercase letter, and (e) the right most recently “and” is Nothing,
Also, (f) the space character to the right of a certain range to the right is ".,
It is not "blank". If this condition is satisfied, the j digit is registered as the end digit of the author name in step 1003. FIG. 11 is a diagram for explaining the title extraction procedure shown in step 704. In step 1100, it is determined whether or not the i digit is the start digit of the title. The condition is (a)
i-1 digit is "" and i digit is character and i digit is leftmost, or (b) i digit is leftmost and i-1 digit is ":" And there is no space to the left of the i digit up to the most recent period character ".", Or (c)
The i digit is more recent to the right of the end digit of the author's name, and i
-1 digit is blank and i is uppercase, or i-1 is "," and i is uppercase, or i-1 is "." And i is uppercase. That is. If the above conditions are satisfied, the i digit is registered as the start digit of the title in step 1101. In step 1102, it is determined whether the j-th digit is the end digit of the title. The condition is (a) that the j-th digit is a character, the j + 1-th digit is ".", And the j + 2-th digit is a blank character. Or (b) j
Digits are letters and j + 1 digits are "." And j + 2
The digits are lowercase, or (c) the j digit is a character and the j + 1 digit is "," and the j + 2 digit is a blank character and the j + 3 digit is an uppercase letter, or (d) j
Digit is character and j + 1 digit is "," and j + 2
The digits are uppercase. If this condition is satisfied, the j digit is registered as the ending digit of the title in step 1103. FIG. 12 is a diagram for explaining the procedure for extracting the magazine name shown in step 705. In step 1200, it is determined whether the i digit is the start digit of the magazine name. The condition is
(A) i-1 digit is "," or (b) i-
One digit is "", or (c) i-1 digit is a blank character, or (d) i-1 digit is ".". If this condition is satisfied, the i digit is registered as the starting digit of the magazine name in step 1201. Also step
1202 determines whether or not the j digit is the end digit of the magazine name. The condition is (a) j + 1 digit is "(", or (b) j + 1 digit is ",". Yes and j + 2
The digit is a lowercase letter, or (c) j + 1 digit is "," and j + 2 digit is a number, or
(D) j + 1 digit is ",", j + 2 digit is a blank character, and j + 3 digit is a lower case letter, or
(E) j + 1 digit is “,”, j + 2 digit is a blank character, and j + 3 digit is a number. If this condition is satisfied, the j digit is registered as the end digit of the magazine name in step 1203. FIG. 13 is a diagram for explaining the procedure of the volume issue extracting process shown in step 706. Now, assuming that the i digit and the j digit are the digits after the end digit of the magazine name, first, in step 1300, it is determined whether or not the i digit is the start digit of the volume issue. (A) i-1 digit is non-numeric and i-digit and i + 1 digit are numeric and i + 2 digit is non-numeric, or (b) i-1 digit is non-numeric and i-digit is Is a number and the i + 1 digit is a non-number, or
(C) i-1 digit is "," and i digit is "v", or (d) i-1 digit is "," and i digit and i + 1 digit are numeric and i + 2 digit is non-numeric, or (e) i-2 digit is "," and i-1
The digit is a blank character and the i digit is "v", or (f) i-2 digit is "," and i-1 digit is a blank character and i digit is a number. ,. If this condition is satisfied, the i digit is registered as the starting digit of the winding number in step 1301. Then, in step 1302, it is determined whether the j-th digit is the end digit of the volume issue.
(A) j digit is a number and j + 1 digit is ","; or (b) j digit is a number and j + 1 digit is a blank character; or (c) j digit is a number. And J
The +1 digit is "(". If this condition is satisfied, the j digit is registered as the end digit of the volume number in step 1303. Fig. 14 shows the page number extraction procedure shown in step 707. In step 1400, it is determined whether or not the i digit is the start digit of the page number, and the condition is (a) i digit and i + 1 digit are “p” and i +
2 digits are “.”, Or (b) i digits and later i + 2
Number up to the digit and i + 3 digit is "-",
Or (c) i digit and i + 1 digit are numbers and i + 2
The digit is “−”, or (d) the i digit is a number and the i + 1 digit is “−”. If this condition is satisfied, the i digit is registered as the start digit of the page number in step 1401. Next, in step 1402, it is determined whether the j digit is the end digit of the page number.
The conditions are (a) j digits are blank characters, or
(B) j digit is ",", (c) j digit is ".", Or (d) j digit is "(", or (e) j- 3 digits are "-" and j-2
Digit from j digit to j digit, or (f) j-2
The digit is “−” and the j−1 digit and the j digit are numbers, or (g) the j−1 digit is “−” and the j digit is a digit. If this condition is satisfied, the j digit is registered as the end digit of the page number in step 1403. FIG. 15 is a diagram for explaining the procedure for extracting the issue year shown in step 708. The condition that the i-th digit is the start digit of the issuance year and the j-th digit is the end digit of the issuance year is as follows. First, (a) i digit is “1”, i + 1 digit is “9”, i + 2 digit and j (= i + 3) digit are numbers, and (b) i−1 digit is a blank character. Or "," or "(" or ".", And
(C) j + 1 digit is a blank character, ",", "(", or ".", Or
(D) i−1 digit is “′”, i digit is a number from 4 to 9 and j (= i + 1) digit is a number,
Is. If this condition is satisfied, in step 1501 the i digit is registered as the starting digit of the issuing year and the j digit is registered as the ending digit of the issuing year. FIG. 16 is a diagram for explaining the procedure for extracting the issue month shown in step 709. First, in step 1600, it is determined whether or not the j digit is the closest character digit to the left of the end digit of the issuing year, and if the condition is satisfied, the j digit is registered as the end digit of the issuing month. Next, in step 1602, it is determined whether the i digit is the start digit of the issuing month. The condition is
(A) The i digit is the closest character digit to the left of the issue month, and (b) the i-1 digit is "," and the i digit is a character, or (c) i The -1 digit is "(" and the i digit is a character. If this condition is satisfied, the i digit is registered as the start digit of the issuing month in step 1603. From FIG. It is a figure showing the result of extracting the bibliographic items.The results of extracting the bibliographic items from the reference documents 1700, 1702, 1704 are shown in 1701, 1703, 1705. Reference numbers, author names, titles, etc. The bibliographic items are extracted Fig. 18 is a diagram for explaining the screen processing procedure of the document database: the search input window frame is displayed in step 1800, and the search result window frame is displayed in step 1801. Enter the search conditions in step 1802, and execute the search according to the search conditions in step 1803. In step 1804, a menu for displaying the search result is selected, and in step 1805, the search result is displayed in a window frame.Figure 19 is a flow chart for generating a variant word taking the author name as an example. First, in step 1900, the author's name is input, and in step 1901, the fast name is extracted, then in step 1902, the middle name is extracted, and in step 1903, the last name is extracted. First, in step 1904, the author names are sorted in the order of fast name, middle name, and last name.
In step 1905, the short name is generated. In step 1906, the author names are sorted in the order of last name, fast name, middle name. In step 1907, the short name is generated. Next, in step 1908, the middle name, last name, and fast name are rearranged in this order, and in step 1909 an abbreviation name is generated. As another method for generating variant words, there is also a method in which a variant word dictionary is provided and a dictionary is searched to retrieve the entry word and its variant word.

【The invention's effect】

According to the present invention, since the notation of the author name and the institution name is different depending on the document and the document has a different notation, it is possible to extract bibliographic items from the paper and generate atypical variants of the bibliographic items. This has the effect of preventing omissions in the search of the literature database. Further, since the title, author name, issue date, etc. can be automatically extracted from the reference of the document, there is an effect that the relation between the document can be automatically registered and the registration work can be saved.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention, FIG. 2 is an explanatory diagram of a procedure for creating a search command for document retrieval, FIG. 3 is an explanatory diagram of variant generation of an institution name, and FIG. Is an explanatory diagram of the variant word generation of the author name, FIG. 5 is a flow chart of registration of mutual relations between documents, FIG. 6 is a flow chart showing a search procedure of a document database, and FIG. 7 is a flow chart of extracting bibliographic items from a reference ,
FIG. 8 is a flow chart of blank character deletion processing, FIG. 9 is a flow chart of reference number extraction, FIG. 10 is a flow chart of author name extraction, and FIG.
Figure 11 is a flow chart for extracting titles, Figure 12 is a flow chart for extracting magazine names, Figure 13 is a flow chart for extracting volume numbers, Figure 14 is a flow chart for extracting page numbers, and Figure 15 is a publication year extraction. Fig. 16, Fig. 16 is a flow chart for extracting the issue month, Fig. 17 is an explanatory diagram of the analysis result of the reference document, Fig. 18 is an explanatory diagram of the screen processing procedure of the document database, and Fig. 19 is a variant generation. It is a flow chart. Explanation of code 100 …… Image input section, 101 …… Bibliographic item extraction section, 103 ……
Reference file, 201 ... variant generation step, 501 ... character extraction and character recognition step, 700 ... reference section input step, 702 ... reference number extraction step, 703 ... author name extraction step, 704 ... Title extraction step, 705 ...
… Magazine name extraction step, 706 …… Volume issue extraction step, 7
07 …… Page number extraction step, 708 …… Issue year extraction step, 709 …… Issue month extraction step, 1700 …… Reference data, 1701 …… Bibliographic item extraction data.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Hiromichi Fujisawa 1-280, Higashi-Kengokubo, Kokubunji, Tokyo (56) References, Hitachi, Ltd. Central Research Laboratory (56) Reference JP-A-1-191258 (JP, A) JP-A-62 -11932 (JP, A) JP 62-216074 (JP, A) JP 63-153630 (JP, A) Proc. Of the 41st National Conference of the Information Processing Society of Japan (second half of 1990) 4-168-169 Operations Research vol. 31, No. 3 (1986) P. 152-157 Tetsuro Ito "Information Retrieval" (Sho 61-8-10) Shokoido P. 15-54

Claims (1)

[Claims] 1. Bibliographic information including the author name, title, journal name, and volume of the reference document by recognizing characters in the reference column described in the printed document and analyzing the character string in the reference column. A method for registering bibliographic data that extracts and registers a series of character strings in the bibliographical column according to the appearance of specific characters, symbols, and character types based on the notation rules in the bibliographical column. In advance, specify the separation rule for separation, enter the above-mentioned bibliography field, detect a blank character from the entered bibliographic field, and if there is no separator before or after the blank character, enter the blank character. , The first digit is the parenthesis symbol or number as the start digit, the parenthesis symbol or the delimiter is the end digit, and the reference number is extracted. The character closest to the right of the end digit of the extracted reference number is the start digit, Parentheses or delimiters Issue of extracting the author's name on the left side of the character as the end of digits, extracted the right side of the character of a by delimiters in the right side of the author's name,
Or, the title is extracted with the right uppercase letter of the author name as the first digit and the left side of the delimiter as the end digit, and the character to the right of the extracted title and to the right of the delimiter is the first digit, and the bracket symbol Or, the journal name is extracted with the left digit of the separator as the end digit, the initial letter or number indicating the volume number on the right of the extracted magazine name as the start digit, and the left digit of the bracket symbol or the separator as the end digit. As the starting digit and the left digit of the parenthesis or the separator as the ending digit. Extract the number, detect the number string that includes the preset number at the beginning to extract the issue year, and detect the character string of the month name that is on the left of the issue year to extract the issue month. Extracted document number, author name, title, journal name, Number, page number, year of publication and month of publication are registered, and in the bibliographic information, from the most recent character to the right of the end digit of the reference number to the character to the left of the parenthesis symbol or separator, A method for registering literature data, characterized in that a variant of at least one of rearranging the order of names and generating an abbreviation is applied to generate an atypical word of an author's name, and is registered by a plurality of notations.